I noticed an error about boinc needing 100MB for a a task and i had run out of space (10GB)
Mon 03 Jul 2017 15:58:55 BST | Einstein@Home | Continuous Gravitational Wave search Galactic Center highFreq needs 84.59MB more disk space. You currently have 15.41 MB available and it needs 100.00 MB.
Mon 03 Jul 2017 15:58:55 BST | Einstein@Home | Gamma-ray pulsar binary search #1 needs 3.67MB more disk space. You currently have 15.41 MB available and it needs 19.07 MB.
So this seemed a lot and i figured it needed to download an updated application, as the tasks were not that large.
So a trundled over to look at who were eating all the disk pies...
3832104 Jul 12 2015 einstein_S5R6_1.01_graphics_i686-pc-linux-gnu
6278871 Feb 2 2016 hsgamma_FGRP4_1.15_x86_64-pc-linux-gnu__FGRP4-SSE2
6320021 Feb 2 2016 hsgamma_FGRPB1_1.00_x86_64-pc-linux-gnu
6320021 Sep 13 2016 hsgamma_FGRPB1_1.01_x86_64-pc-linux-gnu__FGRPOLD
6401891 Dec 9 2016 hsgamma_FGRPB1G_1.05_x86_64-pc-linux-gnu__FGRPSSE
6401891 Sep 26 2016 hsgamma_FGRPB1_1.05_x86_64-pc-linux-gnu__FGRPSSE-Beta
6464399 Jun 21 13:09 hsgamma_FGRPB1_1.06_x86_64-pc-linux-gnu__FGRPSSE
8109487 Oct 4 2016 einsteinbinary_BRP4G_1.39_x86_64-pc-linux-gnu__BRP4G-opencl-ati
8123333 Aug 31 2015 einsteinbinary_BRP4G_1.52_x86_64-pc-linux-gnu__BRP4G-Beta-opencl-ati
8123333 Jun 4 2015 einsteinbinary_BRP6_1.53_x86_64-pc-linux-gnu__BRP6-opencl-ati
10199802 Dec 20 2016 hsgamma_FGRPB1G_1.17_x86_64-pc-linux-gnu__FGRPopencl-ati
10216922 Feb 8 10:06 hsgamma_FGRPB1G_1.18_x86_64-pc-linux-gnu__FGRPopencl1K-ati
10216922 Jan 16 09:02 hsgamma_FGRPB1G_1.18_x86_64-pc-linux-gnu__FGRPopencl-Beta-ati
12272855 Aug 31 2015 einsteinbinary_BRP4_1.00_graphics_i686-pc-linux-gnu
23463592 Jul 12 2015 einstein_S6BucketFU2UB_1.01_x86_64-pc-linux-gnu__X64
23463592 Sep 8 2015 einstein_S6BucketFU3UB_1.01_x86_64-pc-linux-gnu__X64
28296618 Feb 14 2016 einstein_O1AS20-100T_1.02_i686-pc-linux-gnu__SSE2
29826080 Feb 13 2016 einstein_O1AS20-100T_1.02_x86_64-pc-linux-gnu
29827040 Feb 16 2016 einstein_O1AS20-100T_1.03_x86_64-pc-linux-gnu__X64
66192587 Oct 25 2016 einstein_O1MD1TCV_1.03_x86_64-pc-linux-gnu__AVX
66197129 Nov 23 2016 einstein_O1MD1CV_1.00_x86_64-pc-linux-gnu__AVX
66205256 Jun 1 11:22 einstein_O1Spot1TLo_1.04_x86_64-pc-linux-gnu__AVX
66205256 Jun 7 13:00 einstein_O1Spot1Hi_1.00_x86_64-pc-linux-gnu__AVX
66205256 May 19 20:20 einstein_O1Spot1THi_1.04_x86_64-pc-linux-gnu__AVX
72444186 Feb 19 2016 einstein_O1AS20-100T_1.04_x86_64-pc-linux-gnu__AVX
72444186 Mar 9 2016 einstein_O1AS20-100F_1.04_x86_64-pc-linux-gnu__AVXO1F
So can some of these be deleted?
Does the project have a mechanism for freeing these up ?
edit: apologies for the formatting - it looks perfectly fine in the edit window (FireFox)! (press reply to see it formatted correctly)
Copyright © 2024 Einstein@Home. All rights reserved.
I assume those that aren't
)
I assume those that aren't recently mentioned here (app and exact version number) are ancient apps or versions already and are not needed anymore:
https://einsteinathome.org/apps.php
https://einsteinathome.org/server_status.html
https://einsteinathome.org/apps.php?xml=1
strings... BRP4G, BRP6, S6Bucket, O1AS20, O1MD1CV, O1Spot1T, and every 'Beta'
I think you could delete all of those.
Richie_9 wrote:strings...
)
If you do, your BOINC client will probably request them all back again.
This issue comes up from time to time. I remember Bernd saying something to the effect that it was difficult from the server side to 'unsticky' old apps that were no longer current as there wasn't a suitable BOINC mechanism - or something along those lines - I don't remember the details.
If I want to clean up old apps, I remove the entries for them from the state file before deleting them on disk. That way, BOINC 'forgets' about them and doesn't download replacements. I rarely do this as it's easier just to allocate a bit more disk space that BOINC can use. I'm not advocating that people should tamper with the state file unless they really know what they are doing. I just want to make sure people understand that just deleting the file won't solve the problem.
EDIT: The 'official' way to retrieve the disk space would be to set NNT and finish off the tasks on hand. After returning the completed results, resetting the project would get rid of all the old entries. The 'cost' would be that all 'current' files would need to be downloaded afresh. This might be quite a lot with the current GW search. You could avoid most of this if you had taken a backup copy of current files beforehand and restored them before BOINC tried to download them again.
Cheers,
Gary.
This is a BOINC wide problem.
)
This is a BOINC wide problem. There is currently no way for a project to tell the Client that applications can be deleted. Other than manually fiddling with client_state, a project reset is the best way after setting no new work. As Gary said this means that all GW related datafiles will also be downloaded again. You could backup the files but that would still involve client_state fiddling.
I tried removing a few files
)
I tried removing a few files and sure enough on boinc restart it restored them.
edit : i wish i had found xmlcopyeditor earlier, it does make looking at xml files a little easier.
I just use KDE's kwrite for
)
I just use KDE's kwrite for all shell scripts and BOINC xml file editing. I find it very quick and easy for the things I want to do.
Cheers,
Gary.
Christian Beer wrote:This is
)
In fact there is no way for the server to know which application files the client has. Instead IMHO the client would need to be modified to fetch the list of application (Web RPC applications page), then remove the application versions from its internal application list (that client_state.xml is an external representation of) that don't appear there and all application files that are not used in current (read: appearing on the applications page) applications the client already downloaded.
BM
Bernd Machenschalk wrote:In
)
I was thinking the client might be modified to keep track of when application/file was last used, and then remove it when asked to "cleanup old stuff". This would give the client flexibility, as project admins might want to keep the apps available for reruns well into the future, but clients might need to recover the space.
I did toy with the idea of using the file system dates as being representative of an app being used, but the a full system backup kicked that idea into the long grass.